Scientists have unveiled an AI model capable of determining whether any cancerous brain tumour remains during surgery within 10 seconds.
The innovation could revolutionise neurosurgery outcomes.
The technology, named FastGlioma, surpasses traditional methods of identifying residual tumour tissue with remarkable precision.
Spearheaded by researchers from the University of Michigan and the University of California, San Francisco, it demonstrated a significant advantage over current practices.
Senior author Todd Hollon, a neurosurgeon at the University of Michigan Health and assistant professor at U-M Medical School, said: ‘FastGlioma is an artificial intelligence-based diagnostic system that has the potential to change the field of neurosurgery by immediately improving comprehensive management of patients with diffuse gliomas.’
He explained: ‘The technology works faster and more accurately than current standard of care methods for tumour detection and could be generalised to other paediatric and adult brain tumour diagnoses. It could serve as a foundational model for guiding brain tumour surgery.’
During brain tumour surgery, neurosurgeons often face challenges in entirely removing the mass. Residual tumour can resemble healthy brain tissue, making detection challenging. Traditional methods like MRI imaging or fluorescent dyes have limitations, including accessibility and applicability.
To address these challenges, FastGlioma was tested on fresh, unprocessed tissue from 220 patients undergoing surgery for low- and high-grade diffuse gliomas.
The AI model achieved a detection accuracy of around 92%. Remarkably, when comparing surgeries assisted by FastGlioma to those using standard imaging and fluorescence-guided methods, the AI missed high-risk residual tumours only 3.8% of the time, while conventional approaches had a 25% miss rate.
Co-senior author Shawn Hervey-Jumper, professor of neurosurgery at UCSF, said: ‘This model is an innovative departure from existing surgical techniques by rapidly identifying tumour infiltration at microscopic resolution using AI, greatly reducing the risk of missing residual tumour in the area where a glioma is resected.’
He added: ‘The development of FastGlioma can minimise the reliance on radiographic imaging, contrast enhancement or fluorescent labels to achieve maximal tumour removal.’
FastGlioma employs advanced optical imaging combined with foundation AI models. These models, similar to GPT-4 and DALL·E, are trained on vast datasets and can adapt to various tasks. Researchers trained the system on over 11,000 surgical specimens and four million microscopic views using stimulated Raman histology – a rapid, high-resolution imaging method.
The technology offers two imaging modes: a fast mode for lower resolution in 10 seconds and a full-resolution option in 100 seconds. The full-resolution images achieved 92% accuracy, while the quicker mode followed at 90%.
Dr Hollon explained: ‘This means we can detect tumour infiltration in seconds with extremely high accuracy, which could inform surgeons if more resection is needed during an operation.’
Despite advancements in neurosurgery, residual tumour rates have remained static over the past two decades. This issue worsens patient outcomes and adds pressure to global health systems, which anticipate 45 million annual surgeries by 2030.
Global cancer initiatives emphasise integrating innovative technologies like AI into surgical practices. FastGlioma exemplifies such innovation, providing an accessible and cost-effective tool for identifying residual tumours, even in paediatric cases such as medulloblastoma, ependymoma and meningioma.
Co-author Aditya S Pandey, chair of neurosurgery at U-M Health, added: ‘These results demonstrate the advantage of visual foundation models such as FastGlioma for medical AI applications and the potential to generalise to other human cancers without requiring extensive model retraining or fine-tuning.’
Future research aims to extend FastGlioma’s capabilities to other cancers, including lung, prostate, breast, and head and neck tumours
